A few days ago I published a letter from David Levy documenting the collapse of negotiations between the Rybka and Junior camps, failing to agree on the rules for a match between the programs in September. Sergey Abramov, general manager of Convekta, was in charge of negotiations for the Rybka side. He sends a response to Levy's letter and I give it below in full in two parts. It sounds like he is still looking for a match for Rybka. Speaking of, Rybka is currently taking on GM Joel Benjamin in a handicap match, giving the GM a pawn in each game. I have some comments from both sides but ran into time trouble with visiting family this week. I'll have it up soon.
First let us brief on the story of the matter. Early in June Elista hosted the computer match Deep Fritz – Deep Junior. Though the friendly nature of the match was evident (two representatives of the same chess computer company took part), it was widely spread that the winner will have the right to challenge the FIDE World Champion. In such a way the friendly match suddenly turned into qualification for a rather prestigious competition. Although Rybka showed interest, it was not allowed to participate in this selection, in spite of the fact that this program sits firmly at the top of the rating-lists of all independent testing organizations. The result was Vasik Rajlich’s Open Letter, in which he challenged the winner of the match Deep Fritz – Deep Junior.
The rest continues at the link below. Negotiations aside, I don't agree with his statement that chess players all over the world care about high-level computer matches. Unless they are playing against a top human, that is. It's hard to measure something like interest in a sub-sub-section of a sport as marginal as chess already is. But based on the comments and traffic I've seen over the years, there is a small, vociferous band of computer chess junkies who are passionately interested in comp-comp chess. Trying to make the leap from that to the general public may actually be easier than trying to convince the rest of the chess-playing world to care. But they'll need robots making the moves and perhaps an animated movie or two.
First let us brief on the story of the matter. Early in June Elista hosted the computer match Deep Fritz – Deep Junior. Though the friendly nature of the match was evident (two representatives of the same chess computer company took part), it was widely spread that the winner will have the right to challenge the FIDE World Champion. In such a way the friendly match suddenly turned into qualification for a rather prestigious competition. Although Rybka showed interest, it was not allowed to participate in this selection, in spite of the fact that this program sits firmly at the top of the rating-lists of all independent testing organizations. The result was Vasik Rajlich’s Open Letter, in which he challenged the winner of the match Deep Fritz – Deep Junior.
Presidential board of FIDE (27th of June, Tallinn) supported the idea proposed by Vasik Rajlich and recommended that the match Rybka – Junior would be played during the World Championship in September 2007, Mexico. We also conducted the negotiations with Mexican organizers and they agreed to absorb part of costs for organizing the match. Initially, we had proposed a “winner-takes-all” approach, i.e. each side should deposit USD 100,000 towards the prize fund and the winner would get the entire prize fund. However, we were aware of Shay Bushinsky’s standpoint, expressed by him in Elista, that such a match “bet” is unacceptable for Junior. The Israeli side was ready to play with guaranteed share in the prize fund only. Early in July we sent the challenge to Junior, where this standpoint was taken into consideration – we guaranteed the prize fund of USD 100,000 and suggested sharing it between the winner and the defeated side in proportion 70:30. For almost three weeks we had no reply, finally (19th of July) the negotiations started.
One of the sponsors of the World Championship in Mexico is “Intel”, which was willing to provide two identical state-of-art computers for the match. The engineers had the opportunity to arrive in Mexico 4-5 days before the match with ample time to test the equipment. The other option for the teams was to bring the computers with them. An up-to-date 16-core computer costs 10 thousand dollars or slightly more and weighs no more than 40 kg. It is difficult for us to judge why New-York spendings for Junior team in 2003 were 30-40 thousand dollars. Now, as you can see, everything is much cheaper and simpler. Besides it, several respected competitions feature local play. This list includes: Mainz 960 World Championship, Fritz vs. Kramnik match, and even Junior vs. Kasparov match. To our mind, remote game is needed for open events to not exclude anybody. For invitational events, this issue disappears. However, Junior’s team insisted on remote game only.
Let’s return to Elista again. The match Deep Fritz – Deep Junior had 2 press-conferences with engineers. One of them saw the innocuous, one would think, question to be asked: what chess players consulted the programs and who made necessary adjusting between the games (the rules allow it). However, this question made the representatives of both programs to panic a bit, they were confused and claimed promptly something of that sort – there was no interference in the course of the game and such interference was impossible. Besides it, both teams preferred to keep in strict secrecy the assistants’ names, as if the matter concerned the development of a new nuclear bomb. Where in fact, which post address had the basis computers of Deep Fritz and Deep Junior and who was the operators of the remote computers remained sealed information.
We wished to avoid in every possible way any suspicions of unfair game and any scandals like that one, which broke out after the ending of the second match between Kasparov and Deep Blue. As you remember, Kasparov accused the IBM team that the computer used human’s assistance. IBM dismantled the computer in a hurry; however the computer’s triumphant result went down in history, while IBM stocks, according to experts, improved for 11 billion dollars. However, in this issue we met the wishes of Junior’s team and were ready to carry out the match with remote computers. The only conditions we had was transparency of the way computers made their moves and the checking for “assistant-free thinking”. However, with regret we were convinced that our suggestion of open and comprehensive control was not supported by both our rivals and Mr. Levy. That’s why we had to resume our previous position and insisted on on-site game in Mexico. While it didn’t matter for us, whether both sides bring the computers with themselves (there were no insurmountable financial or organizational obstacles) or opt for using the computers, which would have been kindly provided by Intel in Mexico. Unfortunately, none of these suggestions were acceptable for another side.
In conclusion, we would like to point out that history shows that high-level computer matches grab the attention of chess players all over the world. Additionally, computers have done much to increase the popularity of our ancient and ever young game. That’s why we do hope, that during the World Championship in Mexico, we will see Rybka play a fair and hard-fought match against another program which ranks among the absolutely strongest in the world.
Sergey Abramov,
Convekta Ltd, general manager
I definitely agree that computers and computer chess have done much to promote and, yes, improve chess. I'm just not sure what that has to do with comp-comp chess as an attractive competitive venture. Finding creative ways to involve chess computers in big events is now more challenging than beating the top humans!
"Negotiations aside, I don't agree with his statement that chess players all over the world care about high-level computer matches. Unless they are playing against a top human, that is. It's hard to measure something like interest in a sub-sub-section of a sport as marginal as chess already is."
It's immaterial whether chess is marginal or not. Abramov clearly refers to chess player interest not that of the general public. And I would tend to agree with him.
Personally, I'd quite like to see the match. I think the competition between the programmers and the whole David vs. Goliath aspect (Rajilich vs. ChessBase) would be fairly compelling, especially considering it may have the most accurate chess of any match to date.
I am interested in computer chess. But why would this match be different from the games already organized by the SSDF? Rybka is at the top of the crosstable. Number two is an old version of Rybka. I don't know if it's the free demo version. Number three is Hiarcs. Then Junior.
For me, the last sentence of Sergey Abramov's statement is the most interesting. But is it more than hope? Any competitor who takes the challenge, knows that he will almost certainly have to face a match loss. OTOH, if there is a reasonable guaranteed share of the prize funds, as a kind of appearance fee: Then I can imagine that it could be a good chance to present an engine with at least some strong moves and attractive games, to a wider public, as the possibly "second best."
For example, Shredder has about as many World Champion titles as Junior, and currently has some advantages Rybka itself does not offer yet (own interface, bitbases, permanent position learning).
I think it would be worth considering.
Nothing frnd, really useful this site information. But we need some improvement.....
Thx
Yours friendly,
VTS.,
These chess engine vs. engine matches have their place. This is more true now than in the 1990's, because now the best engine presumably plays chess better than anything else, human or machine.
- - - - - -
These matches become less intersting to me if any of the following is true:
(A) Special expensive hardware is used.
A father with bills to pay could obtain a 2-CPU computer from a local shop for under $3000. That should be the hardware limit.
(B) The two engines are run on non-identical hardware.
(C) It is a little bad unless both computers are physically in the same room.
- - - - - -
OLD QUESTION: Can a human usually beat a computer?
NEW QUESTION: Can a human + computer usually beat an identical copy of the computer?
- - - - - -
Would a human vs. computer match give the human a better or a worse chance of winning if -- if the match were switched to chess960 (where neither has any book knowledge)?
My guess is chess960 would give the computer an even better chance of wiping out the human.
The Chess Tigers' annual chess engine vs. engine chess960 competition in Mainz Germany is one of the most visible in the engine vs. engine calendar.
Later this month Rybka will compete in Mainz for its first time.
Yeah, robots making the moves would be cool.
And AIs doing all the home preparation by themselves, fine tuning of opening books, and re-balancing during the match would even be cooler.
And light sabres! Cooooooool!
But seriously, this is not interesting at all, the results are already out there, the point about programs is that they'll replicate the results. When Magnus plays his father today, we do not know the result beforhand; even with the huge rating difference. With a similar situation for programs, (rating not relationship) we would know beforhand, as we do about this match.
Q
"Rybka is currently taking on GM Joel Benjamin in a handicap match, giving the GM a pawn in each game"
Rybka is the new Paul Morphy.
What computer-chess is lacking is the fact that the computer is not happy after beating it's opponent.
The programmers are happy of course, so this is basicly a race between engineers.
However, it's a different kind of sport - programming a computer, and I'm not sure whether or not it should be handled in the same arena where other chess news are - save the daily dirt :)
I still don't understand why a computer gets to use an opening book. I don't get to in my OTB games. I have to 'calculate' during the opening.]
It's "opening book memory" should be exclusively the games it has played OTB historically, and nothing more. That's all I get to go on.
The next step...
I agree with Mark in part. I think that programs are given too much credit for beating humans when they use our historical knowledge of openings. I suspect that most programs would end up in embarrassing situations against top players if they had to rely on their own calculations. On the other hand, I don't think that we necessarily need to match "pure human play" against "pure computer play." I think that using computers to springboard off of our heuristics can only get us closer to perfect play--and can that be bad?!
"Rybka is the new Paul Morphy."
There is inadvertent truth in this. Judging from comments on Chessgames.com, Rybka has fanboys breathless in praise of its genius, the same way that Morphy and Fischer do. Rybka outputs are copied and pasted as though they were pronouncements of the Delphic oracle. I daresay that no other chess engine inspires this level of weird geek devotion.
Stendec wrote:
"I still don't understand why a computer gets to use an opening book."
For the same reason every player, fom patzer to top GM is allowed to bring his memorized chess knowledge to the table.
Remember the days where Kasparov and Fischer were demolishing opponent after opponent with a combination of supreme playing skill and opening preparation?
Computers are doing the same, only at a slightly higher level. Let's face it: computers play better than humans, and it is NOT a problem as long as we don't insist in competing against them.
Handicapping a machine - by taking away the opening book or any other method - in order to give humans a chance to win is outright cheating.
Would you make a car's tires square in order to claim a "win" racing against them?
irv, your position is popular but I've never understood it. it would be no more of a handicap for Junior to play without an opening book than it is for humans to play without a massive engine to aid their calculation. Man vs. Machine matches are only interesting for their asymmetry. Stripping computers of opening books (and possibly endgame tablebases as well) would give everyone what they want -- programmers get a better challenge, fans see more creative chess, and GMs get to win again. The current setup is a holdover for when chess programs were jokes.
Irv's comment shows a basic misunderstanding of what is at stake, the mystery or truth or problem that the audience hopes to find an answer to, when humans play against computers: which plays better?
Allowing one party access to encyclopedic knowledge has little bearing on that question, indeed, detracts from it.
In other words, no one's out to see which competitor can memorize more; obviously, the computer can. Rather, we want to see which competitor can choose better moves. Choose by thinking or calculating, not simply by looking up in a database.
In practice, I think the issue is pretty moot in terms of openings books, which you guys all seem obsessed with -- simply because humans still have the advantage there (super-GM strength humans supported by teams of seconds, including silicon "seconds", have greater chance of coming up with home improvements, or even finding strong novelties over-the-board, than engines blindly following opening books that are based on past human contests).
But it's starting to be germane in endings, where tablebases have grown to the point where I think they can make a real difference in practical play at top levels.
Stripping computers of opening books will definitely make computer chess even less attractive than it is. Watching two computers battling it out in a topical line (say in poisoned pawn or slav) is quite interesting because it gives you insight what might happen in the next Anand-Ivanchuk or Aronian-Kramnik games. Watching Rybka play 1. Nc3 against Junior five times in a row and then winning anyway 80 moves later I find quite boring.
On the issue of the failed match, I find Junior's insistence on remote play unacceptable and frankly suspicious. $100 000 is serious money in chess world. Playing for that amount remotely is the same as having Kramnik and Topalov play their WC match remotely. Wouldn't you love that? Suspicion and allegations would never stop.
Shams wrote:
"irv, your position is popular but I've never understood it. it would be no more of a handicap for Junior to play without an opening book than it is for humans to play without a massive engine to aid their calculation."
My reasoning is very simple, Shams: I accept chessplaying software as it has always been, and that includes the opening book.
If you, and others like Jon Jacobs think that it is ok to strip chess software of some key ingredients in order to make the battle even, I respect your position. That, however, I consider cheating.
Today, it is the opening book. Tomorrow it will be the endings tables. Then it will be the harware; then, something else...
To put it more "human terms", what you propose is that Kasparov be given 2 pounds of sedatives, so that he can play Jon Jacobs - or me! - on "even terms". The excuse? He brings elements to the table that Jon or I lack: top-level training from a very young age, international chess competition experience, a prodigius memory and great natural talent for the game!
I suspect the calls for "disabling" chess software/hardware for humans to win would not be taking place if computers would not have gotten so strong...
BTW, I'm NOT a fan of computer chess in any form. I don't enjoy playing against computers and I don't think there is much to be gained by humans facing machines in chess. I don't own chess-playing software (mainly because I stopped playing competitive chess a long time ago).
The reality, however, can be ignored: computers play better chess than humans TODAY. Everyone is free to come up with a different explanation for the results, but the facts are that machines beat humans regularly these days - and have been doing so for a while.
The more interesting question (for me at least) is: why do professional chess players feel that they stand to gain something from playing (and losing) losing to computers in public?
BTW, I'm NOT a fan of computer chess in any form. I don't enjoy playing against computers and I don't think there is much to be gained by humans facing machines in chess. I don't own chess-playing software (mainly because I stopped playing competitive chess a long time ago).
The reality, however, can't be ignored: computers play better chess than humans TODAY. Everyone is free to come up with a different explanation for the results, but the facts are that machines beat humans regularly these days - and have been doing so for a while.
The more interesting question (for me at least) is: why do professional chess players feel that they stand to gain something from playing (and losing) to computers in public?
Jon wrote:
"Irv's comment shows a basic misunderstanding of what is at stake, the mystery or truth or problem that the audience hopes to find an answer to, when humans play against computers: which plays better?"
There is no mystery there, Jon.
Computers play better - the results are there for everyone to see.
Now, you can blame the opening book or the processor's calculating speed ot the computer's lack of nerves or whatever you fancy as the reason for the machine's success, but the "mystery" of who plays better is long gone.
Let's put it straight: there have been two significant losses in human vs computer matches, Kasparov to Deep Blue and Kramnik to Deep Fritz. What makes you all guys tell that computers consistently win?
I think that a well-prepared human player will ALWAYS stand a good chance to hold against a computer.
There's an entertaining read by GM Soltis on his regular NYPost column about Rybka playing without its opening book:
http://www.nypost.com/seven/07222007/entertainment/computer_at_odds_with_booked_up_foe_entertainment_andy_soltis.htm
irv,
If you are going to quote me, at least copy and paste from my post and not someone else's.
Jon Jacobs,
I think you've hit the nail on the head: "we want to see which competitor can chose better moves." And, I believe, computers are still handily beaten by *top* CC players who have the time to dig below the tactics and do what humans do best--chose better moves.
really, i do know how to spell "choose". good thing i'm not going to have a spelling contest against a computer...
I agree with irv.
It's no surprise that most people are actually interested in who can 'calculate' the best move subject to (humanlike) limitations on memory and processing power rather as opposed to the arbitrary best move. Realize though that endgame tablebases are the result of perfect calculation - seeing all the possibilities right through to the end. Would anyone feel better if current engines were able to do this in real time rather than consulting a pre-computed tablebase? Somehow I don't think so. The sheer inelegance of brute force computation is disdained because it degrades the mystique of chess.
I would love to watch Deep Junior play Rybka
Deep Junior has a unique sacrificial style
Rybka is extremely strong
all this stuff about Deep Junior cheating is Abramov's stupidity and should not have stopped the match from happening
I would have to agree with Shulman!
On both comments!
Cheers!
- D
www.gambetto.com
I would have to agree with Shulman!
On both comments!
Cheers!
- D
www.gambetto.com
Well said irv and Cynical Gripe.
Computers with their opening books, endgame tablebases and immense calculating prowess are the strongest chess playing entities known to us today. Definitely stronger than even the best GM's. What's more, they keep getting stronger by the hour.
If we want to even the tables, we can clearly take away their opening books or endgame tablebases or stipulate that they run on a processor from the early 90's. That would be the equivalent to when you switch to a "friendly mode" when playing Fritz.
Irv ... thanks for the link. I loved the comment about how Rybka might have had the advantage because Ehlvest was unused to the pawn odds position. ;-)
"Allowing one party access to encyclopedic knowledge has little bearing on that question, indeed, detracts from it."
That includes humans too.
"In other words, no one's out to see which competitor can memorize more; obviously, the computer can. Rather, we want to see which competitor can choose better moves. Choose by thinking or calculating, not simply by looking up in a database."
You'd have to take a GM and migically erase his/her opening knowledge memorized as well in order to make that argument. Start the game by telling the GM they have to "choose" or "calculate" each move instead of oplaying anything they have memorized. Its not possible, and trying to equate the computer's superior ability to memorize moves as unfair is meaningless. The whole point is computers play an overall better package of chess.
If you want to use your standpoint, you should be arguing that man versus machine matches should be played on an extended chess variant with all new piece abilities and random shuffling of the home positions. Then you'd have true "choosing" moves by each side, in which case I'm quite certain the computer will win easily.
An article on human versus machine thinking with reference to the Kasparov - Deep Blue match. Nothing new, but a decent read.
http://www.technologyreview.com/Infotech/19179/page1/
Usually I use the best file searcher- http://megaupload.name/